Berlin, Germany

Paul Petersik-Einöder

Climate Physicist · Data Scientist · Python Developer · Founder

"Simple is better than complex." - Principle #3 from the Zen of Python

About

I am a climate physicist, Python developer, and founder building data-intensive applications across weather, agriculture, and mobility. As co-founder of VineForecast, I developed digital tools for sustainable agriculture with a focus on backend systems, data engineering, and applied machine learning.

My work spans backend development with Django, FastAPI, and PostgreSQL, large-scale weather data processing with xarray and zarr, and the development of mechanistic and AI-based models using NumPy and JAX. More recently, I have also worked on event-driven software systems and integrated AI-assisted development workflows into my engineering practice.

My engineering style is strongly influenced by the Zen of Python, with an emphasis on simplicity, readability, and explicit design. My academic background in climate physics and complex systems includes published research on traffic dynamics and machine-learning-based forecasting of El Niño.

Work

Founder & CRO · VineForecast GmbH

The VineForecast GmbH develops digital tools for agriculture, combining meteorology, agronomy and AI.

VineForecast App

A farm management and disease forecasting app used by wine makers throughout the season to digitize field work and support data-driven decisions around vine health and funcgide reduction.

Visit VineForecast

Cultivision AI

Data-driven harvest forecasting for leafy crops, built from sparse, irregular, and noisy field data. It combines meteorological data, variety embeddings, with advanced AI methods, such as neural ordinary differential equations.

Visit Cultivision

Tech Stack

Backend & Web

Django Django REST PostgreSQL Celery FastAPI

ML & AI

JAX Equinox scikit-learn Keras Jupyter Notebook

Data & Geo

Xarray Pandas NumPy rasterio

Infrastructure & Tools

Linux Docker Kubernetes Git Github Actions

Publications

  1. Petersik, P., Panja, D., & Dijkstra, H. A.

    One-parametric bifurcation analysis of data-driven car-following models.

    Physica D: Nonlinear Phenomena, 427, 133016. (2021) — 10.1016/j.physd.2021.133016

  2. Petersik, P., & Dijkstra, H. A.

    Probabilistic forecasting of El Niño using neural network models.

    Geophysical Research Letters, 47(6). (2020) — 10.1029/2019GL086338

  3. Dijkstra, H. A., Petersik, P., Hernández-García, E., & López, C.

    The application of machine learning techniques to improve El Niño prediction skill.

    Frontiers in Physics, 7, 153. (2019) — 10.3389/fphy.2019.00153

  4. Petersik, P., Salzmann, M., Kretzschmar, J., Cherian, R., Mewes, D., & Quaas, J.

    Subgrid-scale variability in clear-sky relative humidity and forcing by aerosol–radiation interactions in an atmosphere model.

    Atmospheric Chemistry and Physics, 18(12), 8589–8599. (2018) — 10.5194/acp-18-8589-2018

Projects

Blog

Simulations and modeling in Julia for plant epidemics and peloton dynamics.

Read on Medium

vivcPy

A Python library for retrieving passport data from the Vitis International Variety Catalogue (VIVC).

View GitHub

Contact

LinkedIn linkedin.com/in/paul-petersik/
GitHub github.com/pjpetersik
Orcid orcid.org/0000-0002-8321-1131
Medium medium.com/@paul.petersik